The 7 Most Important Things To Know About Cisco's New AI-Focused UCS Server

The C480ML is the second new server model Cisco has rolled out since late May, and the new Nvidia GPU-loaded model is intended for customers seeking to harness the data they create to drive favorable business outcomes.

Artificial Flavor
Cisco Systems has unleashed what could be the most important server in the nearly 10-year history of the UCS line. The C480ML is the second new server model Cisco has rolled out since late May, and the new Nvidia GPU-loaded model is intended for customers seeking to harness the data they create to drive favorable business outcomes.

The C480ML is in part a response to rapid customer growth in the artificial intelligence and machine- learning use case, according to Cisco executives. In the past four years, the company has seen 18X customer growth for those use cases, and those customers are demanding deeper insight and analytics.

Customers who gain that insight and harness those analytics put themselves in a position to take some $1.2 trillion from their competitors over the next few years as a result, Cisco estimates. Along the way, partners can recognize deep, profitable opportunity in the variety of services around those efforts.

The Data Explosion

The C480ML is particularly well suited for Internet of Things use cases, Cisco says. IoT devices, Cisco estimates, stand to generate 277X the data end users generate, and 40 percent of all data may come from IoT devices by the year 2020. The rapid growth in IoT comes as some 30 million new connected devices come online every week. By 2020, each person will generate up to 5 TB of data, the company estimates. Cisco Senior Product Manager Han Yang called this a "data firestorm driving need for AI and ML," adding that Cisco intends to eliminate data silos with "a unified architecture that can scale to the large volume needed for AI and ML."

Opportunity For Partners

In the beginning, the C480ML will be handled by Cisco's largest, most high-profile channel partners, like World Wide Technology, ePlus and Presidio. Cisco channel executives, though, see the system as a significant opportunity for the channel in general as the market matures. "There are a lot of nuances required to deliver this type of solution, from consulting services to strategy sessions to workshops to the ability to deliver and measure outcomes," said Scott Mohr, director of data center and cloud inside Cisco's global partner organization. "This gives us ability with our partners to expand the UCS footprint. We have over 80,000 customers for our partners to go back and touch today." Once touched, those customers are likely to depend on solution providers for life, Mohr said. "Partners take the lead in delivering strategy and assessment workshops to help develop AI and ML life-cycle management," Mohr said. "They have the ability to provide full continuity with customers. It puts them in that advisory services role they want to be in."

Applicable Across Industry Verticals

Brent Collins, global practice director, data center at solution provider giant World Wide Technology, said WWT has seen AI adopted across several verticals, and Cisco research indicates wide applicability across major industries. AI is used in the finance industry for fraud detection, administering cryptocurrencies and cyber-training, for example. It's gaining traction in health care in cancer detection, drug discovery and research. It's also used in manufacturing to cut down on product defects, speed production and reduce downtime. Cisco says AI is also being used in industries from media and entertainment to security and retail.

Meeting The Challenge

The rise of AI and machine learning presents the IT industry with significant challenges that Cisco says it and the C480ML are well positioned to counter. For example, distributed data sources and technologies put businesses at risk for increased complexity and siloed operations. The quickly evolving AI and machine- learning space is also at risk for skills shortages, and the massive amount of data being produced demands new data center architectures.

Key Partnerships And Validated Designs

Cisco is offering validated designs that use the C480ML as a foundation while integrating with software solutions like Google Cloud Platform, Cloudera, Kubernetes, TensorFlow, Docker and Apache Spark to handle Hadoop, containers and AI and ML workloads. Cisco sees the C480ML as a natural extension of existing UCS environments alongside mainstream servers, compute-intensive models like the C4200 and C125, and models like the S3260 intended for data-intensive workloads. The ecosystem is lorded over by Cisco's UCS Manager automation platform and its cloud-based Intersight infrastructure management system.

The Stats

Cisco's last major UCS release was the the high-density, AMD-powered C4200, which was introduced in late May. The C480ML comes loaded for bear with eight Nvidia Tesla V100 data center GPUs, two Intel Xeon processors with up to 28 cores per socket and 24 DDR4 DIMMS and up to 3 TB of memory. The system can also be had with up to 24 drives of SAS/SATA or SSD/HDD storage able to handle 182 TB, up to six NVME drives and the ability to handle up to four 100-Gbit network switches.

Putting It All Together

The ecosystem is governed by Cisco's UCS Manager automation platform and its cloud-based Intersight infrastructure management system. The management platforms are a secret sauce for the C480ML and the wider Cisco AI and machine-learning ecosystem. "The biggest thing with Cisco is that they're doing interesting things with AI and Intersight," WWT's Collins said. "They've got some automation advantages going on. The most important thing is the ecosystem they've developed with good offerings from other partner technology. Those ecosystems are going to be really important as people build out data lakes that people can take advantage of. I love that Cisco is putting a lot more time and attention into the AI/business intelligence/analytics space."